the natural history of pollination and mating in bird-pollinated...

PART OF A SPECIAL ISSUE ON PLANT MATING SYSTEMS

The natural history of pollination and mating in bird-pollinatedBabiana (Iridaceae)

Caroli de Waal1, Bruce Anderson2 and Spencer C. H. Barrett1,*1Department of Ecology and Evolutionary Biology, University of Toronto, 25 Willcocks Street, Toronto, Ontario M5S 3B2,

Canada and 2Department of Botany and Zoology, University of Stellenbosch, Private Bag XI, Matieland, 7602, South Africa* For correspondence. E-mail [email protected]

Received: 13 January 2011 Returned for revision: 2 March 2011 Accepted: 10 May 2011 Published electronically: 10 August 2011

† Background and Aims Floral variation, pollination biology and mating patterns were investigated in sunbird-pollinated Babiana (Iridaceae) species endemic to the Western Cape of South Africa. The group includesseveral taxa with specialized bird perches and it has been proposed that these function to promotecross-pollination.† Methods Pollinator observations were conducted in 12 populations of five taxa (B. ringens subspp. ringens,australis, B. hirsuta, B. avicularis, B. carminea) and geographic variation in morphological traits investigatedin the widespread B. ringens. Experimental pollinations were used to determine the compatibility status, facilityfor autonomous self-pollination and intensity of pollen limitation in six populations of four taxa. Allozymemarkers were employed to investigate mating patterns in four populations of three species.† Key Results Sunbirds were the primary pollinators of the five Babiana taxa investigated. Correlated geographi-cal variation in perch size, flower size and stigma–anther separation was evident among B. ringens populations.Experimental pollinations demonstrated that B. ringens and B. avicularis were self-compatible with variation inlevels of autonomous self-pollination and weak or no pollen limitation of seed set. In contrast, B. hirsuta wasself-incompatible and chronically pollen limited. Estimates of outcrossing rate indicated mixed mating withsubstantial self-fertilization in all species investigated.† Conclusions Despite the possession of specialized bird perches in B. ringens and B. avicularis, these structuresdo not prevent considerable selfing from occurring, probably as a result of autonomous self-pollination. In easternpopulations of B. ringens, smaller flowers and reduced herkogamy appear to be associated with a shift to predo-minant selfing. Relaxed selection on perch function due to increased selfing may explain the increased incidenceof apical flowers in some populations.

Key words: Babiana, bird perch, sunbird pollination, floral variation, mating, pollen limitation, relaxedselection.

INTRODUCTION

Bird pollination (ornithophily) has evolved independently inmany lineages of flowering plants. Approximately 65 familiescontain ornithophilous species with many independent originsof the pollination system, often from bee-pollinated ancestors(Stiles, 1981; Cronk and Ojeda, 2008; Wilson et al., 2007;but see Marten-Rodrıguez et al., 2010). Three bird families(Trochilidae – hummingbirds, Nectariniidae – sunbirds,Meliphagidae – honey-eaters) are regarded as flower special-ists (Proctor et al., 1996), and different feeding strategieswithin these families have selected for contrasting suites ofadaptive traits in ornithophilous plants (Westerkamp, 1990).Hummingbirds primarily hover during nectar feeding,whereas sunbirds and honeyeaters mainly perch on variousvegetative or reproductive structures. A considerable literaturedocuments the floral biology of hummingbird-pollinatedspecies (reviewed in Grant and Grant, 1968; Thomson andWilson, 2008), but less is known about the reproductiveecology and floral evolution of species pollinated by sunbirdsand honeyeaters. Most sunbird-pollinated flowers are exploitedby several species (Skead, 1967; Gill and Wolf, 1978),

although there is evidence for floral specialization to eitherlong-billed malachite sunbirds or short-billed speciesamong sunbird-pollinated plants in South Africa (Geerts andPauw, 2009).

Adaptive radiation of pollination systems in AfricanIridaceae is closely associated with floral divergence andspeciation (Goldblatt and Manning, 2006). Babiana, a genusof animal-pollinated perennial geophytes mainly from thewinter rainfall zone of western South Africa, is one of thelarger genera of Iridaceae consisting of approx. 92 species.Most species have showy flowers and are pollinated byinsects of four different orders: Hymenoptera (mainlyApidae), Diptera (mainly Nemestrinidae), Coleoptera (mainlyScarabaeidae) and Lepidoptera (mainly Noctuidae). Goldblattand Manning (2007a, b) proposed that there have been atleast 14 shifts in pollination system in Babiana, including atleast two transitions from insect to bird pollination. FiveBabiana taxa endemic to southern Africa (B. ringens subsp.ringens, B. ringens subsp. australis, B. hirsuta, B. avicularisand B. carminea; Fig. 1) are pollinated by sunbirds, or havebeen inferred to be sunbird pollinated, based on floralmorphology (Goldblatt and Manning, 2007b).

# The Author 2011. Published by Oxford University Press on behalf of the Annals of Botany Company. All rights reserved.

For Permissions, please email: [email protected]

Annals of Botany 109: 667–679, 2012

doi:10.1093/aob/mcr172, available online at www.aob.oxfordjournals.org

mailto:[email protected]



A remarkable adaptation for bird pollination occurs inB. ringens. In this species the sterile inflorescence axis func-tions as a specialized bird perch for visiting malachite sunbirds(Fig. 2). Marloth (1898, 1915) first recorded sunbird visits toB. ringens flowers and suggested the function of the perch.However, until recently the function of the perch inB. ringens had not been investigated experimentally. Basedon observations and manipulative studies of two populationsof B. ringens, Anderson et al. (2005) proposed that the perchmanipulates the position of sunbirds, during nectar feedingfrom the species’ ground-level flowers, and that this facilitatescontact with sexual organs promoting cross-pollination.

The recently described B. avicularis (Goldblatt andManning, 2010) also possesses a naked inflorescence axisand ground-level flowers, raising the question of whether theinflorescence axis in this species has a similar function tothe perch in B. ringens.

Here, the natural history of pollination and mating in bird-pollinated Babiana species is investigated. The main focus ison B. ringens because this species has the widest distributionand displays considerable geographical variation in reproduc-tive traits, particularly flower size (Goldblatt and Manning,2010). The study had four main objectives. (1) To recordflower visitors and their foraging behaviour in populations ofeach of the five Babiana taxa. No previous pollinator obser-vations had been made for B. avicularis and B. carminea, sowe wanted to confirm their putative status as ornithophilousspecies. (2) To investigate geographic variation in perchlength, flower size, floral-tube length and stigma–anther separ-ation in B. ringens. We wanted to determine whether variationin these traits was associated with differences among popu-lations in sunbird visitation, both in terms of visitation ratesand sunbird species. (3) To evaluate the compatibility status,facility for autonomous self-pollination and intensity ofpollen limitation of populations. Pollinator specialization inflowering plants may come at a cost if visitation rates areunreliable, and we sought to determine the extent to whichspecies were capable of maintaining fertility through self-compatibility and autonomous self-pollination. (4) To quantifymating patterns in populations using genetic markers to assess

FI G. 2. Female malachite sunbird using the specialized bird perch of Babianaringens subsp. ringens to feed from flowers.

A B

C

E

D

FI G. 1. Bird-pollinated species of Babiana (Iridaceae) from the Western CapeProvince, South Africa: (A) B. ringens subsp. ringens; (B) B. ringens subsp.

australis; (C) B. hirsuta; (D) B. carminea; (E) B. avicularis.

De Waal et al. — Pollination and mating in bird-pollinated Babiana668

whether specialized adaptations for bird pollination are associ-ated with cross-pollination and high outcrossing rates.

MATERIALS AND METHODS

Study species and sites

Babiana ringens occurs on sandy flats and lower slopes instrandveld and fynbos vegetation. Populations are rarelylarge in size and are typically composed of scattered individ-uals that appear after fire (Table 1). Two subspecies havebeen described differing in geographical distribution and mor-phology (Goldblatt and Manning, 2010). Babiana ringenssubsp. ringens (Fig. 1A) is patchily distributed along theWest Coast region, and individuals have robust perches andlarge red flowers. In contrast, B. ringens subsp. australis(Fig. 1B) occurs on the southern Cape Peninsula and east tothe region surrounding Albertinia. Individuals of this subspe-cies have smaller perches and flowers.

Babiana hirsuta (Fig. 1C) occurs on coastal sand flats anddunes along the west coast of South Africa (Goldblatt andManning, 2007b). Although the species is distributed overapprox. 500 km, it is rarely found .200 m from the shoreline.Babiana hirsuta is clonal with genets producing many showyinflorescences and hundreds of flowers and unlike the otherspecies considered here population sizes can be very large.

The remaining two Babiana species are highly restrictedin distribution and poorly known. Despite extensive fieldworkonly a single population of each species was located. Babianaavicularis (Fig. 1E) is restricted to the west coast and nearinterior of the Western Cape and is reported from just fourlocalities in strandveld (Goldblatt and Manning, 2010).Flowers of B. avicularis are orange to orange–green incolour and are produced on one to three horizontal basalbranches. Babiana carminea (Fig. 1D) is limited to limestoneoutcrops in the Knersvlakte of southern Namaqualand whereplants grow in crevices, and corms are tightly wedged inbedrock. The species is morphologically distinct from the pre-ceding three species and possesses large carmine red flowers

that arise from a tuft of leaves near ground level (Goldblattand Manning, 2007b).

The locations of all populations investigated in the presentstudy are mapped in Fig. 3 and details of localities are listedin Supplementary Data Table S1 (available online). With theexception of B. hirsuta, which was commonly encountered,all populations of the remaining taxa located were includedin this study.

Pollinator observations

Pollinator observations were made during July–October2009 at three populations of Babiana ringens subsp. ringens(HOPE, MAM and ROND), four populations of B. ringenssubsp. australis (SCAR, DEK, VIC1 and VIC2), three popu-lations of B. hirsuta (VELD, LAM1 and ELA), and one popu-lation each of B. avicularis (DRAAI) and B. carminea(VANR). Flower visitors and their behaviour were observedover a period of 1–3 d per site (mean ¼ 2.6 d) for between30 and 190 min each day (mean ¼ 131.1 min d21) just aftersunrise when sunbirds are most active. Observers took thesame vantage point each day so as to view the maximumnumber of plants (mean 31.6 plants per population, range8–80). The observers were far enough from plants so as notto deter birds. For each population the number and speciesof sunbirds visiting Babiana individuals were recorded. Avisit was regarded as when a bird landed on a plant andprobed flowers. Mean visitation rate per population was stan-dardized as the number of visits per inflorescence observedper hour of observations over n number of observation days.Where possible it was also recorded whether sunbirds visitedflowers legitimately, or whether they robbed nectar, and alsoif they were present in the surrounding plant community andnot feeding on Babiana, either flying over the study popu-lation, or sitting and/or feeding on neighbouring vegetation.Finally, in all populations the number of open flowers per indi-vidual showing signs of nectar robbery, in the form of a narrowslit on the floral tube, was recorded.

TABLE 1. Estimated population sizes and observations of sunbirds in 11 populations of five bird-pollinated Babiana taxa (B. ringenssubsp. ringens, B. ringens subsp. australis, B. hirsuta, B. avicularis, B. carminea) during August–October 2009 in the Western Cape

Province of South Africa

Taxon Population

Estimatedpopulation

size

No. ofdays

observed

Totalobservationperiod (min)

Mean no. ofinflorescences

observed per day

Mean visitation rate(visits

inflorescence21 h21) (s.e.)

No. of visits to singleinflorescence

MalachiteSouthern

double-collared

B. ringenssubsp. ringens

MAM 150 3 330 33 0.25 (0.11) 41 0ROND 40 2 240 11 0.20 (0.09) 9 0HOPE 75 3 360 9 0.15 (0.15) 7 0

B. ringenssubsp.australis

SCAR 150 3 290 64 0.13 (0.03) 36 0DEK 50 2 105 15 0.00 0 0VIC1 500 3 540 30 0.004 (0.004) 0 1VIC2 300 1 180 15 0.00 0 0

B. hirsuta VELD .5000 3 460 44 0.65 (0.06) 6 220ELA .1000 3 330 26 0.54 (0.14) 42 36LAM1 .1000 3 430 40 1.08 (0.9) 20 50

B. avicularis DRAAI 300 3 360 24 0.07 (0.07) 0 14B. carminea VANR 90 2 300 18 0.08 (0.03) 8 0

De Waal et al. — Pollination and mating in bird-pollinated Babiana 669

http://aob.oxfordjournals.org/cgi/content/full/mcr172/DC1

Variation in flower visitors were compared among populationsusing log-transformed visitation rate with an ANOVA model thatincluded plant species and populations nested within plantspecies. Unless otherwise noted, this and all subsequent analyseswere performed using JMP Version 8.02 (SAS, 2009).

Variation in morphological traits

To investigate intra-specific variation of perch length, tepalsize, floral-tube length and stigma–anther separation inB. ringens, individuals were randomly sampled from a totalof ten populations (four, B. ringens subsp. ringens; six,B. ringens subsp. australis) during peak flowering fromAugust to October 2009. These traits were also measured inthe single population of B. avicularis. In both species, thenumbers of buds, flowers or fruits on the apical portion ofperches were also recorded. For floral traits, one to threeflowers per individual were sampled using calipers, and themean used in subsequent analyses. Sample sizes for the fourtraits are given in Fig. 4. One-way ANOVAs were used toanalyse morphological variation among populations andTukey–Kramer tests to determine which populations differedsignificantly for a particular trait.

Factors influencing fertility

To investigate factors influencing fruit and seed set, controlledpollinations and caging treatments were performed in one

population of B. ringens subsp. ringens (MAM), two popu-lations of B. ringens subsp. australis (SCAR and VIC1), twopopulations of B. hirsuta (LAM2 and VELD) and one popu-lation of B. avicularis (DRAAI) during July–September2009. Not all types of pollination data were collected fromeach population because of sample size limitations, antelopegrazing and vandalism. At best the following five treatmentswere conducted in each population: (1) hand self-pollination;(2) hand cross-pollination; (3) unmanipulated and caged; (4)removal of stamens; (5) open-pollination. When fruits wereripe, fruit and seed set per fruit were recorded in each treat-ment. Sample sizes are provided in Supplementary DataTable S2. Comparison of the fertility of (1) versus (2) providesevidence for self-compatibility versus self-incompatibility.Treatment (3) provides evidence for the occurrence of auton-omous self-pollination. Treatment (4) indicates the amountof cross-pollination. Comparison of (2) versus (5) providesevidence for pollen limitation of seed set.

For cross-pollinations, all flowers on a plant were pollinatedwith a mixture of pollen from flowers of at least three pollendonors collected several metres away to reduce the incidenceof inbreeding. In B. hirsuta populations, plants were coveredwith mesh wire cages following hand-pollination to preventdamage from antelope herbivory. To investigate whetherplants were self-compatible, flowers were pollinated with self-pollen by rubbing dehisced anthers picked from either thesame flower, or other flowers on the same plant. To investigate

N

0 40 80 160 km

STIL

VIC1

WOR

WORCESTER

WESTERN CAPE

B. avicularisB. carmineaB. hirsutaB. ringens subsp. ringensB. ringens subsp. australis

CAPE TOWN

REDJONK

SCAR

ROND MAM

HOPE

VELD

DRAAI

VANR

CLANWILLIAM

VANRHYNSDORP

LAM2LAM1

ELA

NORTHERN CAPE

VIC2

DEKALBERTINIA

FI G. 3. Localities of bird-pollinated Babiana species from the Western Cape Province, South Africa that were investigated in this study.



the contribution of cross-pollen to fruit and seed set, self-pollination was prevented by removing undehisced anthersfrom mature buds with fine forceps the day before the bud wasdue to open, and this procedure was replicated in all buds onthe inflorescence. It was only possible to emasculate flowers intwo populations of Babiana ringens subsp. australis. To inves-tigate the capacity for autonomous self-pollination, plantswere covered with wire mesh cages before flowering toprevent sunbird visitation and the plants left undisturbed untilfruits were mature. For populations of both B. ringens subspeciesand B. avicularis, hand-pollinations were performed on three orfour consecutive days, pollinating all newly opened flowers andany remaining small buds were removed after this period. InB. hirsuta, where floral displays were too large for supplementalpollinations of all flowers, a subset of flowers was pollinated andthe base of the flowers marked with colour paint so that theycould be identified at harvest.

The effects of the experimental treatments on proportionfruit set and seed set per fruit were analysed using ANOVA.Because not all experimental treatments were replicated ineach population of B. ringens, variation among treatmentswas analysed separately in each population. The same exper-imental treatments were replicated in B. hirsuta populationsand therefore population, treatment and their interaction wereincluded in the ANOVA model. Tukey–Kramer HSD testswere used to examine whether treatment means differed sig-nificantly among B. hirsuta populations, but separatelywithin each population of B. ringens and B. avicularis. Totest for self-compatibility, autofertility and pollen limitation,proportion fruit and seed set per fruit were compared inspecific contrasts of treatments, as outlined above usingTukey–Kramer tests.

Estimates of mating patterns

During July–October 2009, open-pollinated capsules wererandomly sampled at maturity from plants in two populationsof Babiana ringens subsp. ringens (HOPE, WOR), twopopulations of B. ringens subsp. australis (SCAR, VIC1),two populations of B. hirsuta (VELD, ELA) and one popu-lation each of B. avicularis (DRAAI) and B. carminea(VANR). Seeds were ground in a 0.02 M Na2HPO4 buffer(pH 7.4) with DL-dithiothreitol (1 mg mL21) and PVP-40.Extracts were adsorbed onto filter paper wicks and storedat –80 8C until electrophoresis was performed. Wicks werethen inserted into 11–12 % starch gels. Enzymes wereresolved on two buffer systems: (1) glutamate dehydrogenase,triose phosphate isomerase, alcohol dehydrogenase andphosphoglucoisomerase on lithium borate (pH 8.3); and (2)6-phosphogluconate dehydrogenase, phosphoglucomutase,isocitrate dehydrogenase, malate dehydrogenase and acidphosphatase on histidine citrate (pH 6.5). Genotypes at atotal of 15 allozyme loci were inferred based on segregationpatterns characteristic of either dimeric or monomeric codo-minant enzymes; however, not all loci were polymorphicwithin populations and three loci (PGD-2, ACP-1 andACP-3) were polymorphic but not consistently scoreable.

3·5

3·0

2·5

4·5

0

50

0

60

250

200

150

100

50

0

50

40

30

20

10

45403530252015105

a a

F9,356 = 7·22P < 0·0001

F9,361 = 378·49P < 0·0001

F9,361 = 590·65P < 0·0001

F10,413 = 145·72P < 0·0001

a ab abc abc abc c cbc

a a a b bc c e e fd

a ab b c d e e e fe

a a a b b c c d cde ecd

4·0

Mea

n st

igm

a–an

ther

sep

arat

ion

(mm

)M

ean

tube

leng

th (

mm

)M

ean

dors

al te

pal l

engt

h (m

m)

Mea

n pe

rch

leng

th (

mm

)

2·0

1·5

0·5

1·0

050 30 50 50 5049 15935 28

50 30 50 32 509 505015 35

50 30 50 50 915 505035 32

43 25 50 35 1350 9 504927 13

HOPE

RONDM

AMSCAR

RED

DRAAIVIC

2STIL

VIC1

DEK

MAM

ROND

HOPEDEK

SCARRED

STILVIC

1VIC

2

DRAAI

MAM

ROND

HOPE

SCARRED

VIC2

DEkVIC

1STIL

DRAAI

RONDM

AMHOPE

WOR

DEK

SCARVIC

2VIC

1RED

STIL

DRAAI

D

C

B

A

B. ringens subsp. ringensB. ringens subsp. australisB. avicularis

FI G. 4. Mean (+ s.e.) of (A) perch length, (B) dorsal tepal length, (C) tubelength and (D) stigma–anther separation in populations of Babiana ringenssubsp. ringens and B. ringens subsp. australis, and one B. avicularis popu-lation, as indicated. Means that share the same letter are not significantly

different (P , 0.05). Sample sizes are provided for each trait.


The multilocus outcrossing rate (t) was estimated for eachpopulation using the computer program MLTR (version 3.4;Ritland, 1990). This program uses maximum-likelihood toinfer the genotypes of maternal parents, allele frequencies inthe pollen pool, and the proportion of progeny resulting fromoutcrossing. The standard deviation of 1000 bootstrap valueswas used to derive the standard errors of outcrossing rate esti-mates, and the seed family was used as the unit of resampling.Expectation-maximization iteration was used to estimatemaximum-likelihood values of the outcrossing rate. Betweentwo and six loci (mean ¼ four) were used to estimate meanoutcrossing rates in each population.

RESULTS

Sunbird visitation to flowers

During 65.4 h of field observations over 31 d in 12 populationsof five Babiana taxa a total of 492 visits was recorded to indi-vidual inflorescences, of which 34.35 % were made by mala-chite sunbirds, and 65.24 % by southern double-collaredsunbirds. Two visits by a cape sugarbird and a dusky sunbirdwere also recorded in population LAM1 of B. hirsuta. Smallbees and flies were occasionally observed on flowers butthey did not contact the reproductive organs and consequentlythey were not regarded as significant pollinators and are notconsidered further.

Sunbirds differed in their foraging behaviour depending onthe species of Babiana. In the three taxa with specializedperches, sunbirds always alighted on the inflorescence axis,rotated so that their bodies were facing downwards, and thenprobed flowers (Fig. 2). In B. hirsuta, sunbirds commonlylanded on the prominent bend in the inflorescence axis (seeFig. 1C) and then proceeded to probe flowers head down bymoving along the inflorescence axis and by perching on thesturdy side branches. Owing to the diminutive stature ofB. carminea and the absence of perching structures, visitingmalachite sunbirds probed flowers from the ground with thetop of their heads contacting sexual organs.

Visitation rates by sunbirds varied considerably from day today and among Babiana species and populations. Speciesaccounted for a significant fraction of the observed variation invisitation rate (F4,19 ¼ 3.55, P ¼ 0.0251). Two populations ofB. ringens subsp. australis (VIC2 and DEK) received no visitsand VIC1 received only a single visit by a southern double-collared sunbird (Table 1). All three B. ringens subsp. ringenspopulations (MAM, ROND, HOPE), B. ringens subsp. australispopulation SCAR, and the B. carminea population (VANR) werevisited only by malachite sunbirds, whereas the B. avicularispopulation (DRAAI) was visited only by southern double-collared sunbirds. Both malachite and southern double-collaredsunbirds were observed visiting flowers in all three B. hirsutapopulations (VELD, ELA, LAM1).

Although 82 % of all visits to B. hirsuta flowers were madeby southern double-collared sunbirds, most were not legitimateas these birds did not commonly make contact with the sexualorgans of flowers. In contrast, malachite sunbirds probedflowers legitimately. Southern double-collared sunbirds posi-tioned their heads to the side of the flower and gained accessto nectar by making a distinctive slit at the bottom of the

floral tube. The mean percentage of flowers (+ s.e.) perplant that showed signs of nectar robbery in the three popu-lations of B. hirsuta were VELD 91.6+ 5.0, n ¼ 11; LAM144.5+ 4.4, n ¼ 41; ELA 60.3+ 4.8, n ¼ 50. There wasalso evidence of moderate levels of nectar robbery inB. avicularis (DRAAI 21.2+ 4.5, n ¼ 49) and limitedamounts in populations of B. ringens subsp. ringens (HOPE3.1+ 1.71, n ¼ 50) and B. ringens subsp. australis (DEK 0,n ¼ 14; RED 0, n ¼ 4; STIL 11.1+ 11.1, n ¼ 9; VIC11.9+ 1.5, n ¼ 47; VIC2 3.1+ 1.7, n ¼ 50).

Geographic variation in perch and flower traits

Measurements of perch length, flower size, floral-tubelength and stigma–anther separation in 11 populations ofBabiana ringens and B. avicularis revealed significant differ-ences among populations. Three of the four B. ringenssubsp. ringens populations (MAM, ROND and HOPE) hadsignificantly larger perches than populations of B. ringenssubsp. australis and B. avicularis (Fig. 4A). The fourth popu-lation of B. ringens subsp. ringens (WOR) was situated farthersouth-east (Fig. 3) and in this population perches were not sig-nificantly longer than in the B. ringens subsp. australis popu-lation DEK (P ¼ 0.99). Babiana avicularis (DRAAI) had thesmallest perches (mean ¼ 92.1 mm, n ¼ 50), less than halfthe size of those in B. ringens subsp. ringens populations,although not significantly smaller than perches in the STILpopulation of B. ringens subsp. australis (P ¼ 0.34). Sunbirdvisitation rates were positively correlated with perch lengthin the seven populations of B. ringens from which data onthese parameters (R2 ¼ 0.62, F ¼ 8.11, P ¼ 0.046) werecollected.

Flower size also varied considerably over the geographicalrange of B. ringens. MAM, ROND and HOPE had significantlylarger mean dorsal tepal lengths than all other B. ringens popu-lations (Fig. 4B). Babiana avicularis (DRAAI) had significantlysmaller flowers than all populations of B. ringens. Similar pat-terns were also evident for floral-tube length with the shortestfloral tubes evident in eastern populations of B. ringens(Fig. 4C). Associated with this geographic variation in flowersize of B. ringens was variation in the degree of herkogamyamong populations. Populations of subsp. ringens on averageexhibited significantly larger stigma–anther separations thanpopulations of subsp. australis (Fig. 4D).

Apical flowers in Babiana ringens and B. avicularis populations

No plants were observed in the four B. ringens subsp.ringens populations (MAM, HOPE, ROND, WOR) withbuds, flowers or fruits on the apical portions of perches.However, 6.63 % (n ¼ 392) of the individuals of B. ringenssubsp. australis, sampled in five of the six study populations(DEK, RED, STIL, VIC1, VIC2), produced either one ortwo functional flowers on the tips of perches (Fig. 5A, B).Although B. avicularis most commonly produced one to twoside branches at ground level (mean ¼ 1.67, n ¼ 138 plants),plants were found in which the side branches were producedfarther above ground level near the inflorescence terminusresulting in apical flowers, which set fruit as in B. ringenssubsp. australis (Fig. 5C).


Compatibility

Babiana ringens. Fruit and seed set did not differ significantlybetween hand cross- and self-pollination in VIC1, indicatingstrong self-compatibility [Tukey HSD; fruit set, P ¼ 0.35(Fig. 6A); seed set per fruit, P ¼ 0.72 (Fig. 6B)].Self-compatibility was also evident in MAM and SCARbecause, in both populations, fruit and seed set was evidentin caged flowers.

Babiana hirsuta. Hand self-pollination resulted in only 19 %and 23 % of the fruits and 57 % and 71 % fewer seeds perfruit compared with supplemental cross-pollination in VELDand LAM2, respectively [fruit set: VELD, P , 0.0001;LAM2, P , 0.0001 (Fig. 7A); seed set: VELD, P ¼ 0.023;LAM2, P , 0.0001 (Fig. 7B)] indicating moderate self-incompatibility in both populations.

Babiana avicularis. Hand self- and cross-pollinations ofplants at DRAAI produced abundant fruits and values werenot significantly different (P ¼ 0.909) indicating strongself-compatibility (Fig. 8).

Autonomous self-pollination

Babiana ringens. There was variation in the capacity for auton-omous self-pollination among populations of B. ringens. Seedset was evident in caged flowers in populations where it wastested (MAM, SCAR and VIC1), although the amountsobtained were quite variable (Fig. 6B). The fruit set ofcaged flowers was significantly lower than open-pollinatedflowers at MAM and SCAR (MAM, P , 0.0001; SCAR,P ¼ 0.0291), and seed set per fruit was significantly lower inthe autonomous selfing treatment compared with the open-pollinated treatment at MAM, but only marginally significantat SCAR (MAM, P ¼ 0.0068; SCAR, P ¼ 0.0715).However, at VIC1 there was no significant difference in fruitand seed set per fruit between hand self-pollinated andcaged flowers (fruit set, P ¼ 0.19; seed set, P ¼ 0.79), indicat-ing a well-developed capacity for autonomous self-pollination.

Babiana hirsuta. Caged flowers set only 3 % and 6 % of thefruits obtained by hand cross-pollination at VELD andLAM2, respectively. Compared with flowers that had beenself-pollinated, fruit set of caged flowers in both populationswas significantly lower, although in both cases only lowlevels of fruit set were evident (VELD, P ¼ 0.01; LAM2,P ¼ 0.018; Fig. 7A). Seed set per fruit did not differ signifi-cantly between hand self-pollinated and caged flowers ineither population (VELD, P ¼ 0.525; LAM2, P ¼ 0.140;Fig. 7B).

Babiana avicularis. Caged flowers of B. avicularis set abundantfruit, demonstrating a high capacity for autonomous seed set atDRAAI. Caged flowers set 21 % less fruit and produced 38 %fewer seeds than was obtained from hand self-pollination[fruit set, P ¼ 0.01 (Fig. 8A; seed set per fruit, P , 0.0001(Fig. 8B)]. Fruit and seed set of caged flowers was 80 %and 84 %, respectively, of the fruit and seed set obtainedby hand cross-pollination (fruit set, P ¼ 0.03; seed set,P , 0.0001).

A

B

C

FI G. 5. Formation of apical flowers and fruits (A and B, respectively) inBabiana ringens subsp. australis and (C) B. avicularis.


Pollen limitation

Babiana ringens. The two populations of B. ringens investigatedfor pollen limitation of fruit and seed set differed in theirresponse to the treatments (Fig. 6A, B). In MAM, fruit andseed set were 13 % and 18 % higher, respectively, in hand cross-pollinated flowers compared with open-pollinated flowers,demonstrating pollen limitation (fruit set, P ¼ 0.0341; seed setper fruit, P ¼ 0.0023). In contrast, at VIC1 the two treatmentswere not significantly different, indicating no pollen limitationof maternal fertility (fruit set, P ¼ 0.0887; seed set per fruit,P ¼ 0.707).

Babiana hirsuta. There was clear evidence of strong pollen limit-ation of fertility in both populations of B. hirsuta. Averagingacross both populations, open-pollinated fruit set was only 8 %of that obtained from hand cross-pollination. At VELD, fruitset was significantly higher in the supplementary cross-pollination treatment than in the open-pollinated treatment, butthere was no significant difference in seed set per fruit [fruitset, P , 0.0001 (Fig. 7A); seeds per fruit, P ¼ 0.341(Fig. 7B)]. At LAM2 the supplementary cross-pollination treat-ment resulted in significantly higher fruit and seed set (fruit set,P , 0.0001; seed set per fruit, P , 0.0001).

Babiana avicularis. There was no evidence of pollen limitationat DRAAI, as fruit and seed set per fruit did not differsignificantly between the supplemental cross-pollinated andopen-pollinated treatments [fruit set, P ¼ 0.50 (Fig. 8A);seed set per fruit, P ¼ 0.762 (Fig. 8B)].

Contribution of autogamy to self-fertilization

At the SCAR population of B. ringens ssp. australis, fruitset did not differ significantly between the emasculation andopen-pollinated treatment, but seed set per fruit was 34 %lower in the emasculation treatment [fruit set, P ¼ 0.448(Fig. 6A); seed set per fruit, P ¼ 0.0055 (Fig. 6B)]. AtVIC1, both fruit set and seed set per fruit were significantlylower in the emasculation treatment compared with theopen-pollinated treatment (fruit set, P , 0.0001; seed set,P ¼ 0.0029).

Mating patterns

Of the eight populations of bird-pollinated Babiana speciesthat were screened for allozyme polymorphisms, only fourpopulations of three species provided sufficient variation to

0

2

4

6

8

10

12

14

16b a aa abb ab a abb a

OP HCP OPASP ASPEM ASP OP HSPEM HCP

B

See

d se

t per

frui

t

Treatment

0

0·2

0·4

0·6

0·8

1·0

b a ac ab

MAM SCAR VIC1

ab b b ac ab

A

Pro

port

ion

frui

t set

1·2

FI G. 6. Mean (+ s.e.) of (A) proportion fruit set and (B) seed set in populations of Babiana ringens subsp. ringens (MAM) and B. ringens subsp. australis (SCARand VIC1) following experimental treatments: open-pollinated (OP), emasculation (EM), hand self-pollination (HSP), autonomous self-pollination (ASP) and sup-

plementary cross-pollination (HCP). Means that share the same letter are not significantly different (P , 0.05) among treatments within each population.


estimate mating patterns. All populations displayed mixedmating with low to moderate frequencies of outcrossing:B. ringens subsp. ringens HOPE (tm ¼ 0.32+ 0.10; n ¼ 14families/11.4 seeds per family; n ¼ 5 loci); B. hirsuta VELD(tm ¼ 0.44+ 0.19; n ¼ 25 families/11.4 seeds per family;n ¼ 2 loci) and ELA (tm ¼ 0.25+ 0.05; n ¼ 20 families/9.4seeds per family; n ¼ 3 loci), and B. carminea VANR(tm ¼ 0.49+ 0.12; n ¼ 23 families/9.6 seeds per family;n ¼ 6 loci). The remaining populations were monomorphicat most loci, or if they possessed polymorphic loci they exhib-ited strongly unbalanced allele frequencies making estimatesof mating parameters unreliable.

DISCUSSION

This study provides novel information on pollination andmating in four Babiana species endemic to the Cape regionof South Africa. The pollinators of this group appear to beexclusively sunbirds that forage in distinct ways, dependingon the plant species they encounter. Three taxa possessspecialized bird perches that facilitate nectar feeding from

flowers. Correlated geographic variation was found in perchsize and floral traits among B. ringens populations, withsmaller-flowered populations characterized by very lowsunbird visitation. The experimental pollinations demonstratedthat B. ringens and B. avicularis were self-compatible witheither weak or no pollen limitation of seed set. In contrast,B. hirsuta was moderately self-incompatible, with the fertilityof populations strongly pollen limited. The marker-based esti-mates indicated mixed mating, including substantial self-fertilization in all populations that were investigated. Below,the ecological and evolutionary significance of these findingsis discussed and explanations provided for several seeminglyunexpected results obtained that have general implicationsfor mating system inference.

Bird pollination in Babiana

The observations of B. ringens subsp. ringens confirm pre-vious reports (Goldblatt et al., 1999; Anderson et al., 2005)that this taxon is pollinated by malachite sunbirds (Table 1).In contrast, of the four B. ringens subsp. australis populations

A0·7

0·6

0·5

0·4

0·3

Pro

port

ion

frui

t set

0·2

0·1

0

c b c

LAM2

a c b c

VELD

a

B8

7

6

5

4

3

See

d se

t per

frui

t

2

1

0OP HSP ASP HCP OP HSP ASP HCP

b b b a b b b b

Treatment

FI G. 7. Mean (+ s.e.) of (A) proportion fruit set and (B) seed set in populations of Babiana hirsuta (LAM2 and VELD) following experimental pollinationtreatments: open-pollinated (OP), hand self-pollination (HSP), autonomous self-pollination (ASP) and hand cross-pollination (HCP). Means that share the same

letter are not significantly different (P , 0.05) among treatments across populations.


included in the present study, only SCAR was visited by mala-chite sunbirds. In the remaining three eastern populations onlya single southern double-collared sunbird was observed visit-ing flowers in a single population. The lack of visitation inthese populations was not due to an absence of sunbirds inthe habitats they occupied, as malachite, southern double-collared and orange-breasted sunbirds were observed at thesites. At VIC1 and VIC2 large patches of flowering proteac-eous shrubs, most notably Leucospermum praecox, occurredadjacent to Babiana populations and were commonly visitedby sunbirds. Thus, competition for pollinators may explainthe absence of visitation to B. ringens in these populations.The diminutive stature, relatively low rewards and sparsedensity of individuals of this taxon may make populationsespecially vulnerable to competition from larger and moreabundant Proteaceae with abundant nectar rewards.

This is the first study to report sunbird pollination inB. avicularis, making it one of the smallest flowered bird-pollinated species in the Iridaceae (Goldblatt and Manning,2010). Southern double-collared sunbirds used the perches ofthis species when foraging for nectar in a similar way to mala-chite sunbirds visiting B. ringens. It is plausible that thesmaller flowers of B. avicularis (and perhaps also B. ringenssubsp. australis) have resulted, in part, from selectionimposed by smaller sunbird species, particularly the southern

double-collared sunbird. The larger malachite sunbird wasnot observed visiting B. avicularis, although it was presentin the habitat, raising the possibility of pollinator-driven differ-entiation in flower and perch size and the partitioning of polli-nation resources among bird-pollinated Babiana species.

The present study also provides the first direct evidence thatsunbirds pollinate the large-flowered B. carminea (Table 1),confirming the predictions of Goldblatt and Manning (2007b).The morphology and presentation of B. carminea flowers dif-fered markedly from the other bird-pollinated Babianaspecies. Pale markings on the tepals of B. carminea flowers remi-niscent of nectar guides may reflect a recent shift from long-tongued fly to bird pollination (Goldblatt and Manning,2007b). A shift in the opposite direction seems unlikely, asB. carminea is the only bird-pollinated species nested within aclade of insect-pollinated flowers (Schnitzler et al., 2011). It ispossible that an ecological shift by the ancestor of B. carmineato the highly specialized habitats now occupied by this speciesmay have triggered a transition in pollination system. BecauseB. carminea individuals do not have perches, visiting malachitesunbirds probe flowers while on the ground, or when perching onsurrounding rocks, a relatively unusual feeding position for sun-birds, although one that is occasionally observed in Babiana(Goldblatt and Manning, 2007a; see fig. 1C in Anderson et al.,2005) and other bird-pollinated plants (e.g. Calceolaria uni-flora; Sersic and Cocucci, 1996). An elongated bird perch mayhave been difficult to evolve in B. carminea if ancestral popu-lations resembled most long-tongued fly and bee-pollinatedBabiana species, which do not possess flowers supported by asturdy inflorescence.

Southern double-collared sunbirds were observed robbingnectar from flowers of B. hirsuta by making conspicuous slitsat the bottom of the floral tube. Nectar robbing is commonplacein flowering plants, although it is more commonly reported forbee-pollinated species and may have diverse ecological andevolutionary consequences (reviewed in Irwin et al., 2010). Incontrast to southern double-collared sunbirds, malachite sun-birds visiting B. hirsuta probed for nectar in a legitimatemanner. This contrast in feeding behaviour most likely occursbecause of differences between the two species in bill andtongue length (Rebelo, 1987), with the shorter-billed southerndouble-collared sunbird unable to access nectar by probingfrom the mouth of the flower. On average 34 % of the flowersin B. hirsuta populations showed signs of nectar robbing.These results corroborate a recent study of B. hirsuta byGeerts and Pauw (2009) reporting that all visits by southerndouble-collared sunbirds were illegitimate, whereas those bymalachite sunbirds during the same observation period werelegitimate. Nectar robbing may alter the behaviour of malachitesunbirds, as birds commonly avoid nectar-robbed flowers inother systems (Irwin et al., 2010). For a self-incompatiblespecies such as B. hirsuta this could have important reproductiveconsequences, and nectar robbing may have contributed to thevery low female fertility observed (Fig. 7).

Geographic patterns of trait variation in Babiana ringens

The patterns of geographic variation in morphological traitsof B. ringens that were detected largely reflect the taxonomicdifferentiation between subsp. ringens in the western portion

A1·0

0·8

0·6

0·4

Pro

port

ion

frui

t set

0·2

0

ab a b a

B12

8

10

6

4

See

d se

t per

frui

t

2

0OP HSP ASP

Treatment

HCP

b a b b

FI G. 8. Mean (+ s.e.) of (A) proportion fruit set and (B) seed set in a popu-lation of Babiana avicularis (DRAAI) following four experimental treatments:open-pollinated (OP), hand self-pollination (HSP), autonomous self-pollination(ASP) and supplementary cross-pollination (HCP). Means that share the same

letter are not significantly different (P , 0.05) among treatments.


of the range and subsp. australis in the east (Goldblatt andManning, 2010). In general, plants became more diminutive instature moving eastwards and this was associated with correlatedchanges in perch length, flower size, floral-tube length andstigma–anther separation. It seems likely that this variationhas a selective basis, but the mechanisms responsible are notknown. It will be particularly important to determine the relativeimportance of abiotic versus biotic factors governing the originand maintenance of this variation. Less reliable rainfall(Cowling and Lombard, 2002), regional variation in soil ferti-lity, and differences in the intensity of competition for sunbirdpollination resulting from co-occurring flowering plant speciescould all play a role. Interestingly, disjunct populations ofsubsp. australis on the Cape Peninsula (SCAR, RED) had some-what larger flowers than were typical in eastern populations ofthis subspecies (Fig. 4B); but whether this pattern results fromgreater rainfall or more reliable pollination by sunbirds, includ-ing the larger malachite sunbird, or both, is unclear. Commongarden and reciprocal transplant studies would be valuable toinvestigate these problems further.

Several lines of evidence suggest that populations ofB. ringens subsp. australis in which sunbird visitation waslow or absent experience higher rates of selfing comparedwith the larger-flowered B. ringens subsp. ringens. First,stigma–anther separation in subsp. australis is significantlyless than in subsp. ringens (Fig. 4D), facilitating autonomousselfing in the absence of sunbird pollination. Reduced herko-gamy is commonly associated with increased selfing ratesin flowering plants (Barrett and Shore, 1987; Takebayashiet al., 2006). Secondly, observed average heterozygosity insubsp. australis was substantially lower (0.004+ 0.001,n ¼ 7 populations) than in subsp. ringens (0.027+ 0.008,n ¼ 4 populations), based on a survey of 13 allozyme loci(C. de Waal, B. Anderson and S. C. H. Barrett, unpubl. res.).Finally, the failure to suppress apical branch development inperches of B. ringens subsp. australis (Fig. 5) suggests thatnatural selection maintaining the integrity of perches may berelaxed (Lahti et al., 2008), as one might expect in populationsthat have shifted to high rates of selfing. If this is the case, wemay be observing the evolutionary dissolution of the perch as aresult of the failure of B. ringens subsp. australis to competefor sunbird pollination.

Factors influencing female fertility

The results of experimental pollinations confirm a previousreport of self-compatibility in B. ringens subsp. ringens(Anderson et al., 2005). The present results extend thisfinding by demonstrating that B. ringens subsp. australis(Fig. 6) and B. avicularis (Fig. 8) are also highly self-compatible and populations have the ability to autonomouslyself-pollinate to different degrees. Through comparisons ofsupplemental cross- and open-pollinations it was also demon-strated that populations of B. ringens exhibited either weak orno pollen limitation of seed set. Maternal fertility inB. avicularis also showed no evidence of pollen limitation.In these taxa it seems likely that the capacity for autonomousself-pollination may function as a mechanism of reproductiveassurance when sunbird visitation is unreliable.

Controlled pollinations of B. hirsuta revealed moderatelystrong self-incompatibility in both populations, although theextent to which early-acting inbreeding depression mightalso contribute to low seed set following self-pollination isunknown (see Husband and Schemske, 1996). The two popu-lations of B. hirsuta investigated were chronically pollenlimited. Populations can suffer from pollen limitation whenfloral densities are far in excess of the number of effective pol-linators (e.g. Larson and Barrett, 1999). Given the large floraldisplays in B. hirsuta populations it seems likely that this is themain reason for the very low fruit set recorded, as visitation bymalachite sunbirds was substantially lower than by the primar-ily nectar-robbing southern double-collared sunbirds at the twosites studied for pollen limitation (Table 1). In addition, nectarrobbing may also potentially damage the styles of B. hirsutaflowers and negatively affect female function. However, theincidence of pollen limitation can vary significantly in bothspace and time (e.g. Baker et al., 2000) and future populationstudies over several years are desirable before firm conclusionscan be reached. Obtaining this information is of particularimportance because of earlier claims of pervasive pollen limit-ation of fruit set in the Cape Floristic Region (Johnson andBond, 1997).

Self-incompatibility is commonly associated with clonalityin plants (reviewed in Vallejo-Marın et al., 2010). Clonalspecies often exhibit large floral displays resulting in consider-able geitonogamous self-pollination (e.g. Eckert, 2000). Itseems likely that self-incompatibility in B. hirsuta is main-tained in populations to limit fitness costs (e.g. inbreedingdepression and pollen discounting) associated with geitono-gamy. Clones of B. hirsuta often consist of many floweringramets and daily floral displays are orders of magnitudelarger than in the other bird-pollinated Babiana species.Consequently, the extent of geitonogamous self-pollinationseems likely to be much higher in B. hirsuta and theauthors’ own observations indicated that sunbirds frequentlyprobed many flowers on clones before departing. Variationin the pollination environment will, however, have less influ-ence on lifetime fitness of B. hirsuta clones compared withthe non-clonal Babiana species because their greater longevityshould buffer them against annual variation in outcross pollenreceipt.

Cautionary lessons on mating system inferences

It is often assumed that specialized pollination systems func-tion, in part, to promote high outcrossing rates and improveprogeny quality. However, it has recently been recognizedthat the flowers of many specialized pollination systemsoften have the ability to self-pollinate autonomously, presum-ably because of the risks of specialization in an uncertain pol-lination environment (Zhang et al., 2005; Fenster andMarten-Rodrıguez, 2007). The present results on mating pat-terns in B. ringens suggest that this species may representsuch a case. The large showy red flowers and specializedperches of B. ringens subsp. ringens are clearly adaptationsfor promoting cross-pollination by sunbirds. This initially ledus to the naıve expectation of high outcrossing rates in popu-lations. Yet, estimates from three populations of this speciesindicate relatively high levels of self-fertilization in each


(mean s ¼ 0.63; Anderson et al., 2005; this study). Similarly,half of the seeds produced by B. carminea resulted from self-fertilization despite its large showy bird-pollinated flowers. Itseems probable that mating patterns in these species arehighly sensitive to rates of sunbird visitation. In years withreliable visitation, outcrossing may be higher than reportedhere but, when visitation is low, autonomous self-pollinationmay provide reproductive assurance. However, the role thatautonomous self-pollination plays in providing reproductiveassurance will depend on the magnitude of inbreedingdepression. Unfortunately, as is the case in most plantspecies with mixed mating, nothing is known in Babianaabout the relative fitness of selfed and outcrossed offspringproduced by open-pollinated plants.

Populations with chronic pollen limitation pose similar chal-lenges for mating system inference. Controlled self- and cross-pollinations of B. hirsuta demonstrated that this species has amoderately strong self-incompatibility system also leading toan expectation of high outcrossing rates. However, thepresent estimates of mating patterns indicated significantamounts of selfing in both populations (VELD, s ¼ 0.56;ELA, s ¼ 0.75). This apparent contradiction can be reconciledby considering that the open-pollinated fruit set used formating-system analysis represented only 5 % and 4 % of thepotential fruit set for VELD and LAM2, respectively,because of substantial pollen limitation in both populations.Self-incompatibility in B. hirsuta, as in many species, is notabsolute allowing some seed to be produced from self-pollination (Fig. 7). As a result, most of the open-pollinatedfruits used for mating-system estimates probably originatedfrom flowers with pollen loads composed of significantamounts of self-pollen due to the scarcity of cross-pollen depo-sition. Despite the recent burgeoning literature on matingsystems (reviewed in Goodwillie et al., 2005) and pollen limit-ation (reviewed in Harder and Aizen, 2010) in plant popu-lations, most studies on these topics have been conducted inisolation from one another. Future empirical work on the quan-titative relationships between the intensity of pollen limitationand variation in mating patterns are long overdue.

SUPPLEMENTARY DATA

Supplementary data are available online at www.aob.oxford-journals.org and consist of the following. Table S1: specificlocations with GPS coordinates of all Babiana populationsinvestigated in this study. Table S2: samples sizes for theexperimental pollination treatments.

ACKNOWLEDGEMENTS

We thank S. P. and Lauriette de Waal, Darius van Rensburgand Anina Heystek for assistance with field work, JohnManning for helpful advice on Babiana, Bill Cole andMelinda Pickup for assistance with figures and analysis andJanet Naude, Nico de Jager, Niekie de Jager, Marnus Truter,Jakkie van der Westhuizen and staff at Rondeberg PrivateNature Reserve for locality information and permission towork on their farms. C.deW. was supported by a ConnaughtFellowship from the University of Toronto, and the research

was funded by a Discovery Grant from NSERC and CanadaResearch Chair support to S.C.H.B.

LITERATURE CITED

Anderson B, Cole WW, Barrett SCH. 2005. Specialized bird perch aidscross-pollination. Nature 435: 41–42.

Baker AM, Barrett SCH, Thompson JD. 2000. Variation of pollen limitationin the early flowering Mediterranean geophyte Narcissus assoanus(Amaryllidaceae). Oecologia 124: 529–535.

Barrett SCH, Shore JS. 1987. Variation and evolution of breeding systems inthe Turnera ulmifolia complex (Turneraceae). Evolution 41: 340–354.

Cowling RM, Lombard AT. 2002. Heterogeneity, speciation/extinctionhistory and climate: explaining regional plant diversity patterns in theCape Floristic Region. Diversity and Distributions 8: 163–179.

Cronk Q, Ojeda I. 2008. Bird-pollinated flowers in an evolutionary and mol-ecular context. Journal of Experimental Botany 59: 715–727.

Eckert CG. 2000. Contributions of autogamy and geitonogamy to self-fertilization in a mass-flowering, clonal plant. Ecology 81: 532–542.

Fenster CB, Marten-Rodrıguez S. 2007. Reproductive assurance and theevolution of pollination specialization. International Journal of PlantSciences 168: 215–228.

Geerts S, Pauw A. 2009. Hyper-specialization for long-billed bird pollinationin a guild of South African plants: the malachite sunbird pollination syn-drome. South African Journal of Botany 75: 699–706.

Gill FB, Wolf LL. 1978. Comparative foraging efficiencies of some montanesunbirds in Kenya. Condor 80: 391–400.

Goldblatt P, Manning JC. 2006. Radiation of pollination systems in theIridaceae of sub-Saharan Africa. Annals of Botany 97: 317–344.

Goldblatt P, Manning JC. 2007a. Floral biology of Babiana (Iridaceae:Crocoideae): adaptive floral radiation and pollination. Annals of theMissouri Botanical Garden 94: 709–733.

Goldblatt P, Manning JC. 2007b. A revision of the southern African genusBabiana, Iridaceae: Crocoideae. Section 2.2 Antholyzoides. Strelitzia18. South African National Biodiversity Institute, Pretoria and MissouriBotanical Garden, Missouri.

Goldblatt P, Manning JC. 2010. New taxa of Babiana (Iridaceae:Crocoideae) from coastal Western Cape, South Africa. Bothalia 40:47–53.

Goldblatt P, Bernhardt P, Manning JC. 1999. Evidence of bird pollinationin the Iridaceae of southern Africa. Adansonia 3: 25–40.

Goodwillie C, Kalisz S, Eckert CG. 2005. The evolutionary enigma of mixedmating in plants: occurrence, theoretical explanations, and empirical evi-dence. Annual Review of Ecology, Evolution and Systematics 36: 47–79.

Grant V, Grant KA. 1968. Hummingbirds and their flowers. New York, NY:Columbia University Press.

Harder LD, Aizen MA. 2010. Floral adaptation and diversification underpollen limitation. Philosophical Transactions of the Royal Society ofLondon Series B 365: 529–543.

Husband BC, Schemske DW. 1996. Evolution of the magnitude and timing ofinbreeding depression in plants. Evolution 50: 54–70.

Irwin RE, Bronstein JL, Manson JS, Richardson L. 2010. Nectar robbing:ecological and evolutionary perspectives. Annual Review of Ecology,Evolution and Systematics 41: 271–292.

Johnson SD, Bond WJ. 1997. Evidence for widespread pollen limitation offruiting success in Cape wildflowers. Oecologia 109: 530–534.

Lahti DC, Johnson NA, Ajie BC et al. 2008. Relaxed selection in the wild.Trends in Ecology & Evolution 24: 487–496.

Larson BMH, Barrett SCH. 1999. The ecology of pollen limitation in buzz-pollinated Rhexia virginica (Melastomataceae). Journal of Ecology 87:371–381.

Marloth R. 1898. Die Ornithophilie in der Flora Sudafrikas. Berichte derDeutschen Botanischen Gesellschaft 19: 176–179.

Marloth R. 1915. Flora of South Africa. Cape Town: Darter Brothers.Marten-Rodrıguez S, Fenster CB, Agnarsson I, Skog LE, Zimmer EA.

2010. Evolutionary breakdown of pollination specialization in aCaribbean plant radiation. New Phytologist 188: 403–417.

Proctor M, Yeo P, Lack A. 1996. The natural history of pollination. Portland,OR: Timber Press.

Rebelo AG. 1987. Bird pollination in the Cape Flora. In: Rebelo AG. ed. Apreliminary synthesis of pollination biology in the Cape Flora. Pretoria:CSIR, 83–108.







Ritland K. 1990. A series of FORTRAN computer programs for estimatingplant mating systems. Journal of Heredity 81: 235–237.

SAS. 2009. JMP, Version 8.02. 1989–2009. SAS Institute Inc, Cary, NC.Schnitzler J, Barraclough TG, Boatwright JS et al. 2011. Causes of plant

diversification in the Cape biodiversity hotspot of South Africa.Systematic Biology 60: 343–357.

Sersic AN, Cocucci AA. 1996. A remarkable case of ornithophily inCalceolaria: food bodies as reward for a non-nectarivorous bird.Botanica Acta 109: 172–176.

Stiles FG. 1981. Geographical aspects of bird–flower coevolution, with par-ticular reference to Central America. Annals of the Missouri BotanicalGarden 68: 323–351.

Skead CJ. 1967. The sunbirds of Southern Africa. Cape Town, South Africa:Cape Transvaal Printers.

Takebayashi N, Wolf DE, Delph LF. 2006. Effect of variation in herkogamyon outcrossing within a population of Gilia achilleifolia. Heredity 96:159–165.

Thomson JD, Wilson P. 2008. Explaining evolutionary shiftsbetween bee and hummingbird pollination: convergence, divergence,and directionality. International Journal of Plant Sciences 169:23–28.

Vallejo-Marın M, Dorken ME, Barrett SCH. 2010. The ecological andevolutionary consequences of clonality for plant mating. Annual Reviewof Ecology, Evolution and Systematics 41: 193–213.

Westerkamp C. 1990. Bird-flowers: hovering versus perching exploitation.Botanica Acta 103: 366–371.

Wilson P, Wolfe AD, Armbruster WS, Thomson JD. 2007. Constrainedlability in floral evolution: counting convergent origins of humming-bird pollination in Penstemon and Keckiella. New Phytologist 176:883–890.

Zhang L, Barrett SCH, Gao JY et al. 2005. Predicting mating patterns frompollination syndromes: the case of “sapromyiophily” in Tacca chantrieri(Taccaceae). American Journal of Botany 92: 517–524.


the natural history of pollination and mating in bird-pollinated...

Documents